Optimal Feature Selection for Radar Signal Classification with Different Targets Situation

The detection of the target depends on the accuracy of the classification of the radar return signals. This classification accuracy is based on the features which extracted from that signal. This paper presents an optimal algorithm for select optimal features. Three cases were studied of the situation of targets in receiving signal; single, multi-, and close multi-targets. The types and number of features represent the base of the algorithm, while the processing time and the classification rate represent the criteria for features selection. Hybrid methods, which combine the optimum characteristics of wrappers and filter methods are used for making the compromise between the number of features and best candidate subset. Multilayer perceptron back propagation neural network has been used as a classifier, while the classification is 98% for single target with two third processing time of multi-targets for the same classification rate, and nearly half processing time for the close-multi target.